2026 Baseline · Wave 1 · Data collected February–July 2026
Your best producers are losing their week to work nobody hired them for, and for the first time, we have their own numbers on it.
Key findings
- 75% of surveyed insurance agents use ChatGPT for work, and 34% use ChatGPT and nothing else.
- 53% run their AI on free plans. Another 10% are not sure whether they are paying at all.
- Average self-rated AI confidence is 2.65 out of 5, just below the scale's midpoint. Nearly half (45%) rate themselves a 1 or 2.
- Free-plan users are more than twice as likely as paid users to put themselves at the bottom of that scale (54% vs 21%).
- 43% name CRM updating and data entry as the task that feels least like selling.
- 72% repeatedly look up the same product, plan, or compliance information while talking with a client.
The morning this data describes
One agent in this survey described her routine like this: log into every insurance company portal, then re-log into HealthSherpa and healthcare.gov every 30 minutes, all day. Her own accounting of the cost was "a minimum of 30 minutes a day." Another described entering client information into Medicare Center, then re-entering the same information a second time just to run a quote.
Before every insurance workshop I run, I send the room a short survey about how they work and where the week actually goes. Between February and July of 2026, more than a hundred insurance professionals, most of them Medicare-focused agents and agency staff, filled it out ahead of three industry events, including two trainings hosted by FMOs. The exact count and how the survey works are in the methodology section at the bottom, so you can judge the data for yourself.
Keep one thing in mind as you read: every person in this sample signed up for AI training. These are the motivated ones. That makes what they told me more sobering, not less, and I'll come back to why.
Adoption is wide and shallow
Nearly everyone in the sample has touched AI. Three quarters (75%) use ChatGPT for work. Gemini reaches 31%, Microsoft Copilot 30%, and Claude 15%. Only about 5% said they don't use AI at all.
Look one layer deeper and the picture changes. A third of the sample (34%) uses ChatGPT and nothing else. Put plainly: one agent in three has only ever seen one AI tool, so "AI" to them means whatever ChatGPT happens to do.
The plans they're on tell you even more. More than half (53%) are running free versions, and another 10% couldn't say whether they were paying at all. When a professional can't tell you whether their tool is the free version, that tool hasn't become part of how they work. It's still a website they visit sometimes.
This is what Pilot Purgatory looks like at the individual level. Agencies experience it as pilots that never scale. Agents experience it as a free chatbot tab, opened occasionally, asked to polish an email. One respondent put it plainly: "I don't know what I could ask it to beyond polishing my emails."
The confidence gap
We asked every respondent how comfortable they are using AI to assist with their daily sales tasks, on a scale of 1 to 5, where 1 means "not at all" and 5 means "it's my primary assistant." Throughout this piece I'll call that their AI confidence: how ready they feel to actually put AI to work on the job.
The average landed at 2.65. On a five-point scale, a 3 is the neutral middle, so the average agent sits just below neutral. And the average hides the real story: 45% rated themselves a 1 or 2, while just 9% chose 5. Nearly half the sample is still at the starting line.
The split behind that average is the finding I keep coming back to. Agents on paid plans averaged 3.31. Agents on free plans averaged 2.36. Among free users, 54% put themselves at the bottom of the scale, compared with 21% of paid users. In plain terms: agents who invested in an AI tool mostly describe themselves as getting somewhere with it, and agents dabbling on free versions mostly describe themselves as stuck.
Free-plan agents are more than twice as likely to rate their AI confidence at the bottom of the scale: 54% vs 21% for paid users.
To be fair to the data, a survey like this can't prove which causes which. Maybe paying for better tools builds skill, or maybe the agents who were already serious are the ones who chose to pay. Either way, the pattern is worth acting on: free-tier dabbling and confidence do not travel together.
Why the real industry numbers are probably worse
Here's the part I promised to come back to, and it's worth slowing down for.
Everyone in this sample chose to register for AI training. Think about who that leaves out: the agents who are skeptical of AI, the ones intimidated by it, and the ones too buried in their book of business to even consider a workshop. Those agents, the ones most likely to be struggling with AI, mostly aren't in this data, because people don't sign up to train on something they're avoiding.
So this survey captures the curious and the motivated end of the industry. And even among that group, nearly half rated their AI confidence at the bottom of the scale.
If the most motivated slice of the industry is at 45%, the share across all agents, where the reluctant far outnumber the curious, is almost certainly higher. I can't put a precise number on the whole industry from this data, and I won't pretend to. What I can say is that every number in this piece is best read as the optimistic version of the story.
The work they were never hired to do
We asked what task in a typical week feels like data entry rather than selling. The answers clustered hard. Out of everyone surveyed, 43% named CRM updating and data entry, 30% named email and follow-up, and 12% named logging meeting notes. Nearly half the room, unprompted, pointed at the same task.
Read the individual answers and you see what those categories actually mean. One agent's data entry was "putting in the names of doctors and medicines." Another closes a sale, then types everything she just discussed into her CRM by hand. Nothing in that work requires her judgment or her license. It is transcription, moving what she already knows from her head into software fields, and every hour of it is an hour of Admin Drag: time a skilled professional spends working as a typist.
The compliance side of the week tells the same story. Asked what feels like a time-waster but is required anyway, 24% described hunting for product specs and information, 20% named compliance and marketing review, and 12% named building custom proposals from scratch.
The information hunt
One survey question asked whether there's a specific product, service, or case study detail agents find themselves repeatedly looking up, or asking a teammate about, while talking with a prospect. Nearly three out of four (72%) named one. OEP rules. Medicaid details. What's actually covered in a summary of benefits. Ancillary products they don't feel fluent in yet.
72% of agents repeatedly look up the same product, plan, or compliance information while talking with a client.
Stop and picture that stat in a workday: a client on the phone, a question the agent has answered a dozen times before, and the conversation pausing again while the agent goes digging for the same answer in a portal, a PDF, or a coworker's memory.
The survey wasn't designed to measure the cost of jumping between systems, and I want to be straight about that. But the answers volunteered it anyway: the portal re-logins, the double entry, the mid-call hunts. Harvard Business Review calls this cost the Toggle Tax, and the agents who counted it counted in daily half-hours.
Almost every one of those hunts ends at information the agent's own organization already has. The knowledge exists. It just doesn't show up where the conversation is happening.
What they would do with ten hours back
The survey's last big question: if you could reclaim ten hours of your week from administrative work, what's the one revenue-generating activity you'd spend it on?
Researching new market territories led at 31%. Deeper relationship building with key accounts came in at 21%, and more discovery and prospecting calls at 21%. Nobody, in more than a hundred responses, said they'd use the time to work less.
That's the stat I'd want every agency principal and event planner to sit with. Given room to breathe, these professionals would grow: new markets, deeper client relationships, more conversations. The ambition is fully intact. What's missing is the time, and one response about long-term planning captured where the time goes instead: "Right now all I do is react to the biggest crisis."
When we described a simple alternative, a one-page briefing on every lead delivered each morning, the reactions ranged from "it would help" to "30-45 mins and my day would change drastically."
What I take from all of this
Read one way, the confidence numbers look like an indictment of agents. I read them as an indictment of what agents were handed. The industry gave licensed professionals a free chatbot tab, no redesign of the work underneath, and no governance around any of it, then wondered why confidence stalled below neutral.
The agents in this data don't need another tool recommendation. The work itself has to change shape: the transcription handed off, the mid-call hunts ended, the morning research finished before the morning starts. That's the redesign a Digital Crew exists to deliver, and it's why the free-versus-paid gap matters less as a pricing observation than as evidence of a deeper divide, between professionals dabbling with AI and professionals whose work has been rebuilt around it.
The encouraging part of this dataset is the last section. Give these people their hours back, and they already know exactly where those hours will go.
What to do with these numbers
If you're an agent, the way out of the 45% is smaller than you think. Pick the one task in your week that feels most like transcription, which for most of the room is the CRM, and make it the first thing you learn to hand to AI. One task handled well beats ten experiments abandoned.
If you run an agency, look at the paid-versus-free gap and see it for what it is: a picture of what your producers have been given. Confidence followed investment everywhere in this data. Before your next tooling decision, ask how many of your people are working AI questions out on free plans, and what that costs you in hours they spend typing instead of selling.
If you plan events or education for this industry, this data describes your room. Nearly half the people who choose to show up for AI training arrive at a 1 or 2. Program for the starting line, hands-on sessions where the room builds something they can use Monday morning, and skip the future-of-AI highlight reel.
Methodology
Responses were collected February through July 2026 via a pre-workshop survey administered ahead of three industry events, including two FMO-hosted trainings. 112 insurance professionals responded, which is the sample size, n=112 if you're citing it formally. Respondents were insurance sales professionals, predominantly Medicare-focused agents and agency staff, who had registered for AI training. Because the sample is self-selected and likely more AI-curious than the industry at large, these numbers are best treated as a ceiling, not a floor. AI confidence was self-rated in response to the question "how comfortable are you using AI to assist with your daily sales tasks," on a 1-5 scale anchored at 1 ("not at all") and 5 ("it's my primary assistant"). Open-text responses were categorized by keyword, and a single response can fall into more than one category, so those percentages describe how often a theme was named rather than exclusive shares of the sample. Percentages are rounded to whole numbers. This is Wave 1 of a recurring series, and core questions are frozen so future waves can be compared directly against this baseline.
How to cite this data
Cite as: The AI Hat Insurance AI Adoption Survey, 2026 Baseline (n=112). https://theaihat.com/insurance-ai-adoption-survey/
The charts on this page are free to reuse with attribution and a link back to this page. If you have questions about the data or want to see a cut we didn't publish, reach me through the site.
Frequently asked questions
What AI tools do insurance agents use most?
In The AI Hat's 2026 survey of 112 insurance professionals, ChatGPT led at 75%, followed by Gemini (31%), Microsoft Copilot (30%), and Claude (15%). A third of respondents used only ChatGPT.
How confident are insurance agents with AI?
Average self-rated confidence using AI for daily sales tasks was 2.65 on a 5-point scale. Nearly half (45%) rated themselves a 1 or 2, and only 9% said AI is their primary assistant.
Do insurance agents pay for AI tools?
53% of surveyed agents used free plans, 38% used paid plans, and 10% were unsure whether they were paying. Paid users reported markedly higher confidence (3.31 vs 2.36 average).
What administrative work slows insurance agents down most?
CRM updating and data entry (43%), email and follow-up (30%), and logging meeting notes (12%) topped the list, with hunting for product information the leading compliance-related time cost (24%).
If you're wondering where you sit against these numbers, that's exactly what the AI Adoption Reality Check measures. It takes a few minutes, and it will tell you whether you're in the 45%, and what to do about it.
